Date: Fri, 10 Dec 2004 14:53:43 -0600 From: "James R. Van Artsalen" <james@jrv.org> To: Mike Hunter <mhunter@ack.Berkeley.EDU> Cc: freebsd-amd64@freebsd.org Subject: Re: Tyan Thunder K8W (S2885) with FreeBSD 5.3-RELEASE Message-ID: <41BA0CD7.4080309@jrv.org> In-Reply-To: <20041210172111.GB10811@ack.Berkeley.EDU> References: <20041210.194845.653222828.chat95@mac.com> <20041210172111.GB10811@ack.Berkeley.EDU>
next in thread | previous in thread | raw e-mail | index | archive | help
5.3-release has a known bug with more than 4GB of RAM. You need 5.3-stable. But even that may not work: I have a Tan S2885 also with two Opteron 248s and 8GB, two DIMMs on each CPU. This ran FreeBSD/amd64 5.2.1 since spring with no problems using ATA and bge(4) NIC. I've now switched to a twa(4) 3ware 9500S disk controller, which doesn't need bounce-buffers. bge(4) does need bounce-buffers. FreeBSD/amd64 5-stable hangs right after the bge(4) NIC is enabled if 8GB of RAM is present. I can boot single user and do disk I/O with no problem, but it hangs - no panic - as soon as I try to bring up the NIC. If I add the line "set hw.physmem=4g" to the top of /boot/loader.rc then everything works fine - no bge(4) hang. I think the em(4) NIC doesn't need bounce-buffers and I'm trying to track one down for testing. I think drivers had to be changed for the bounce buffer changes in 5.3. Perhaps there is a bug in the bge(4) change? Is anyone using a bge(4) NIC with more than 4GB of RAM? Mike Hunter wrote: >On Dec 10, "NAKATA Maho" wrote: > >Nakata-san, > > > >>Yesterday, I installed 5.3-RELEASE to my Opteron dual box. >> >>o Opteron 242 (1.6MHz) x2 >>o Memory ECC PC2700 Reg. trancend 2G modules x4 >> + CenturyMicro 512M x 4 >> (total 10G) >>o Tyan Thunder K8W (S2885) >> >>I encountered a prolbem, while installtion. As many one pointed out >>in this ML, if we have >4G of memory, ATA hangs up. >>so I type >>o set hw.physmem="3G" >>at boot time. >>however, I recompiled and re-installed SMP kernel, >>FreeBSD never boots even if I type set hw.physmem="3G". >> >>So I took different way. There's no problem with 5.2.1-RELEASE >>and 5.3-BETA4. So CVSup'ing from 5.3-BETA4 to -STABLE >>seems to be fine, only set hw.physmem="3G" is mandatory at >>installation, and can omit after recompiling world and kernel. >> >>Does anybody here have problems with it? >> >> > >I have a similar story. 5.2.1 loaded fine, 5.3-RELEASE would cause >instant reboots. I was able to cvsup to RELENG_5 and run that kernel. At >the time I blamed bad 5.3 CD Media, but perhaps there was something >actually wrong. I have a Tyan K8S with 8G of ram. > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?41BA0CD7.4080309>